1 Static imbalance no drift

Friedman rank sum test for G-mean
Friedman chi-squared = 47.388, df = 4, p-value = 1.266e-09

OOB UOB OB VFDT ESOS_ELM
1.18 2.35 3 3.94 4.53

Friedman rank sum test for Recall
Friedman chi-squared = 45.129, df = 4, p-value = 3.737e-09

UOB OOB OB ESOS_ELM VFDT
1.12 2.29 3.59 3.65 4.35

2 Imbalance ratio changes

Friedman rank sum test for G-mean
Friedman chi-squared = 159.46, df = 4, p-value < 2.2e-16

OOB UOB OB VFDT ESOS_ELM
1.17 2.25 3.59 3.95 4.03

Friedman rank sum test for Recall
Friedman chi-squared = 182.19, df = 4, p-value < 2.2e-16

UOB OOB ESOS_ELM OB VFDT
1.06 2.3 3.28 4.12 4.23

3 Single factor borderline

Friedman rank sum test for G-mean
Friedman chi-squared = 35.92, df = 4, p-value = 3.005e-07

OB OOB UOB VFDT ESOS_ELM
1.5 1.7 2.8 4 5

Friedman rank sum test for Recall
Friedman chi-squared = 20.56, df = 4, p-value = 0.000387

ESOS_ELM OOB OB UOB VFDT
2.2 2.3 2.4 3.2 4.9

4 Single factor rare

Friedman rank sum test for G-mean
Friedman chi-squared = 22.64, df = 4, p-value = 0.0001494

OOB VFDT OB UOB ESOS_ELM
1.8 2.3 2.4 3.9 4.6

Friedman rank sum test for Recall
Friedman chi-squared = 19.6, df = 4, p-value = 0.0005989

ESOS_ELM VFDT OB OOB UOB
1.6 2 3.7 3.8 3.9

5 Single factor join

Friedman rank sum test for G-mean
Friedman chi-squared = 22.267, df = 4, p-value = 0.0001774

OOB OB UOB VFDT ESOS_ELM
1.33 1.67 3.17 3.83 5

Friedman rank sum test for Recall
Friedman chi-squared = 12.533, df = 4, p-value = 0.0138

OOB UOB OB VFDT ESOS_ELM
2.17 2.17 2.17 4.17 4.33

6 Single factor move

Friedman rank sum test for G-mean
Friedman chi-squared = 23.333, df = 4, p-value = 0.0001086

OOB OB UOB VFDT ESOS_ELM
1 2.17 2.83 4 5

Friedman rank sum test for Recall
Friedman chi-squared = 20.4, df = 4, p-value = 0.0004163

OOB OB UOB VFDT ESOS_ELM
1.17 2 3 4.17 4.67

7 Single factor split

Friedman rank sum test for G-mean
Friedman chi-squared = 22.933, df = 4, p-value = 0.0001306

OOB OB UOB VFDT ESOS_ELM
1.33 1.67 3 4 5

Friedman rank sum test for Recall
Friedman chi-squared = 22.267, df = 4, p-value = 0.0001774

OOB OB UOB VFDT ESOS_ELM
1.33 1.67 3 4.17 4.83

8 Single difficulty factors

Friedman rank sum test for G-mean
Friedman chi-squared = 148.54, df = 4, p-value < 2.2e-16

OOB OB UOB VFDT ESOS_ELM
1.4 2.24 2.93 3.65 4.78

Friedman rank sum test for Recall
Friedman chi-squared = 38.124, df = 4, p-value = 1.057e-07

OOB UOB OB ESOS_ELM VFDT
2.33 2.53 2.85 3.33 3.96

9 Pair imbalance-borderline

Friedman rank sum test for G-mean
Friedman chi-squared = 127.06, df = 4, p-value < 2.2e-16

UOB OOB ESOS_ELM OB VFDT
1.25 2.17 2.72 4.25 4.6

Friedman rank sum test for Recall
Friedman chi-squared = 139.64, df = 4, p-value < 2.2e-16

UOB ESOS_ELM OOB OB VFDT
1.52 1.57 2.9 4.42 4.58

10 Pair imbalance-rare

Friedman rank sum test for G-mean
Friedman chi-squared = 63.44, df = 4, p-value = 5.483e-13

UOB OOB ESOS_ELM VFDT OB
2.02 2.35 2.62 3.52 4.47

Friedman rank sum test for Recall
Friedman chi-squared = 112.68, df = 4, p-value < 2.2e-16

UOB ESOS_ELM OOB VFDT OB
1.52 1.9 3.02 3.83 4.72

11 Pair imbalance-move

Friedman rank sum test for G-mean
Friedman chi-squared = 38.667, df = 4, p-value = 8.163e-08

UOB OOB ESOS_ELM VFDT OB
1.5 2 2.5 4.33 4.67

Friedman rank sum test for Recall
Friedman chi-squared = 42.467, df = 4, p-value = 1.335e-08

UOB ESOS_ELM OOB VFDT OB
1.42 1.67 2.92 4.33 4.67

12 Pair imbalance-join

Friedman rank sum test for G-mean
Friedman chi-squared = 30.667, df = 4, p-value = 3.58e-06

UOB OOB ESOS_ELM VFDT OB
1.5 2 3 4.17 4.33

Friedman rank sum test for Recall
Friedman chi-squared = 42.933, df = 4, p-value = 1.068e-08

UOB ESOS_ELM OOB VFDT OB
1.25 1.83 2.92 4.42 4.58

13 Pair imbalance-split

Friedman rank sum test for G-mean
Friedman chi-squared = 54.222, df = 4, p-value = 4.728e-11

UOB OOB ESOS_ELM VFDT OB
1.39 2.17 2.61 4.22 4.61

Friedman rank sum test for Recall
Friedman chi-squared = 58.933, df = 4, p-value = 4.86e-12

UOB ESOS_ELM OOB VFDT OB
1.39 1.89 2.83 4.17 4.72

14 Pair split-borderline

Friedman rank sum test for G-mean
Friedman chi-squared = 106.73, df = 4, p-value < 2.2e-16

OOB UOB ESOS_ELM VFDT OB
1.82 1.88 3 4.02 4.28

Friedman rank sum test for Recall
Friedman chi-squared = 113.96, df = 4, p-value < 2.2e-16

UOB ESOS_ELM OOB VFDT OB
1.74 2.26 2.46 4.14 4.4

15 Pair split-rare

Friedman rank sum test for G-mean
Friedman chi-squared = 77.248, df = 4, p-value = 6.665e-16

UOB OOB ESOS_ELM VFDT OB
1.92 2.28 2.84 3.64 4.32

Friedman rank sum test for Recall
Friedman chi-squared = 98.992, df = 4, p-value < 2.2e-16

ESOS_ELM UOB OOB VFDT OB
1.92 2.04 2.9 3.54 4.6 
## Warning: Removed 46 rows containing missing values (geom_point).

## Warning: Removed 46 rows containing missing values (geom_point).

## Warning: Removed 46 rows containing missing values (geom_point).

## Warning: Removed 46 rows containing missing values (geom_point).

16 Multiple difficulty factors

Friedman rank sum test for G-mean
Friedman chi-squared = 389.53, df = 4, p-value < 2.2e-16

UOB OOB ESOS_ELM VFDT OB
1.76 2.15 2.91 3.91 4.27

Friedman rank sum test for Recall
Friedman chi-squared = 481.47, df = 4, p-value < 2.2e-16

UOB ESOS_ELM OOB VFDT OB
1.76 1.96 2.82 3.95 4.51

17 All datasets

Friedman rank sum test for G-mean
Friedman chi-squared = 631.08, df = 4, p-value < 2.2e-16

OOB UOB ESOS_ELM OB VFDT
1.85 1.98 3.34 3.89 3.94

Friedman rank sum test for Recall
Friedman chi-squared = 738.8, df = 4, p-value < 2.2e-16

UOB ESOS_ELM OOB VFDT OB
1.71 2.32 2.68 4.06 4.22

18 Real world data

  Dataset Iteration       Range Safe Borderline Outlier Rare   IR

1 Tripadvisor 1 0-2000 0.42 0.33 0.14 0.11 2.31 2 Tripadvisor 2 2000-4000 0.39 0.39 0.13 0.10 2.10 3 Tripadvisor 3 4000-6000 0.37 0.33 0.13 0.17 3.65 4 Tripadvisor 4 6000-8000 0.40 0.35 0.13 0.13 2.37 5 Tripadvisor 5 8000-10000 0.36 0.33 0.16 0.15 2.93 6 Tripadvisor 6 10000-12000 0.45 0.32 0.13 0.09 2.48 Minority.ratio 1 0.3021148 2 0.3225806 3 0.2150538 4 0.2967359 5 0.2544529 6 0.2873563

[1] “Tripadvisor” [1] “Amazon” [1] “Twitter” [1] “PAKDD”

[1] “Tripadvisor”

## `summarise()` regrouping output by 'Example' (override with `.groups` argument)

[1] “Amazon”

## `summarise()` regrouping output by 'Example' (override with `.groups` argument)

[1] “Twitter”

## `summarise()` regrouping output by 'Example' (override with `.groups` argument)

[1] “PAKDD”

## `summarise()` regrouping output by 'Example' (override with `.groups` argument)
## Warning: package 'apcluster' was built under R version 4.0.3
## 
## Attaching package: 'apcluster'
## The following object is masked from 'package:stats':
## 
##     heatmap
## Warning: package 'farff' was built under R version 4.0.3
## Parse with reader=readr : ../../real-streams/streams/amazon.arff
## Loading required package: readr
## header: 0.000000; preproc: 0.050000; data: 0.110000; postproc: 0.000000; total: 0.160000
## Parse with reader=readr : ../../real-streams/streams/twitter.arff
## header: 0.000000; preproc: 0.050000; data: 0.060000; postproc: 0.000000; total: 0.110000
## Parse with reader=readr : ../../real-streams/streams/tripadvisor.arff
## header: 0.010000; preproc: 0.110000; data: 0.130000; postproc: 0.000000; total: 0.250000
## Warning in .local(s, x = x, ...): algorithm did not converge; turn on details
## and call plot() to monitor net similarity. Consider
## increasing 'maxits' and 'convits', and, if oscillations occur
## also increasing damping factor 'lam'.
## Parse with reader=readr : ../../real-streams/streams/pakdd.arff
## header: 0.000000; preproc: 0.080000; data: 0.190000; postproc: 0.000000; total: 0.270000
## Warning in .local(s, x = x, ...): algorithm did not converge; turn on details
## and call plot() to monitor net similarity. Consider
## increasing 'maxits' and 'convits', and, if oscillations occur
## also increasing damping factor 'lam'.

[1] “amazon” [1] “twitter” [1] “tripadvisor” [1] “pakdd”

## Warning: package 'viridis' was built under R version 4.0.3
## Loading required package: viridisLite
## Parse with reader=readr : ../../real-streams/streams/pakdd.arff
## header: 0.000000; preproc: 0.070000; data: 0.180000; postproc: 0.000000; total: 0.250000
## Parse with reader=readr : ../../real-streams/streams/amazon.arff
## header: 0.010000; preproc: 0.040000; data: 0.050000; postproc: 0.000000; total: 0.100000
## Parse with reader=readr : ../../real-streams/streams/twitter.arff
## header: 0.010000; preproc: 0.050000; data: 0.060000; postproc: 0.000000; total: 0.120000
## Parse with reader=readr : ../../real-streams/streams/tripadvisor.arff
## header: 0.000000; preproc: 0.110000; data: 0.130000; postproc: 0.000000; total: 0.240000